646
I Set A Trap To Catch Students Cheating With AI. The Result Was Deflating (discuss.tchncs.de)
submitted 9 months ago by to c/fuck_ai@lemmy.world
276 comments fedilink hide all child comments
 

‘But there is a difference between recognising AI use and proving its use. So I tried an experiment. … I received 122 paper submissions. Of those, the Trojan horse easily identified 33 AI-generated papers. I sent these stats to all the students and gave them the opportunity to admit to using AI before they were locked into failing the class. Another 14 outed themselves. In other words, nearly 39% of the submissions were at least partially written by AI.‘

Article archived: https://web.archive.org/web/20251125225915/https://www.huffingtonpost.co.uk/entry/set-trap-to-catch-students-cheating-ai_uk_691f20d1e4b00ed8a94f4c01

The article‘s author, Will Teague, with his cat, Persephone ’Dots‘ Teague
you are viewing a single comment's thread
view the rest of the comments
[–] 11 points 9 months ago (2 children)

I think a good way to deal with this would be assignments that also (partly) prepare you for the written exam. So if you sit down and do it yourself and actually understand the assignment you already did some learning for the exam.

I had one course at university with homework assignments that were super tough. I did it all by myself but in the end I learned so much that I didn't even need to study for the written exam and got a top grade. Others who did not do their assignments on their own had to study hard for the written exam and in most cases got way worse grades or failed.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 6 points 9 months ago

    Yeah I had a class once where the exams were always just slightly modified versions of some of the homework questions. So if you were confident you could understand and complete all the homework you knew you'd be fine on the exam. It actually caused some people to study more since they felt like there was a more 1:1 between study time and exam success.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 5 points 9 months ago

    I think a good way to deal with this would be assignments that also (partly) prepare you for the written exam. So if you sit down and do it yourself and actually understand the assignment you already did some learning for the exam.

    My networking degree was built on a similar principle. A large part of the degree was configuring Cisco switches and routers, which is done through a command line interface, and you can streamline this by copy/pasting a series of commands. The first test was simply setting the hostname, a password and configuring ssh access. This was a full hour+ test, meanwhile by the end of the semester that was the first five minutes of an hour long test to configure more advanced functionality on these devices

    The entire degree program was setup so that you take the configs you wrote in the first week and keep building onto them and keep copy/pasting bits and pieces of the configs all the way through graduation and into the workforce.

    I could absolutely see a programming degree program taking a similar approach, where you write code snippets in the first class of your freshman year that evolve and you're ultimately still using by the time you graduate. It forces you to know your code, gives you good code to build off of professionally and makes it really difficult to coast by with nasty hacks and/or AI doing it all for you. It also sets the perfect trap for those trying to have AI do everything for them, where they won't know what it's doing or why it's breaking as the underlying snippets are constantly being changed. Especially if it's structured with a lot of easy "plumb your existing snippets together" assignments where it's dead simple for those actually doing the coursework but really hard for those relying on AI entirely. They'll be forced to learn or drop out which is really important for a college experience!

  • permalink
  • fedilink
  • source
  • parent